iiinova — where your digital self keeps evolving

storage.courtlistener.com

The government can't contract around the constitution. The ruling doesn't reach the harder First Amendment questions CDT’s report takes up, but the ruling is pretty clear that the gov’t can’t use procurement to bully potential contractors for their speech. storage.courtlistener.com/recap/gov.us...

Order on Motion for Summary Judgment – #250 in Anthropic PBC v. U.S. Department of War (N.D. Cal., 3:26-cv-01996) – CourtListener.com

ORDER ON 166 215 CROSS MOTIONS FOR SUMMARY JUDGMENT; GRANTING 169 MOTION TO SUPPLEMENT THE ADMINISTRATIVE RECORD. (rfllc3, COURT STAFF) (Filed on 8/27/2026) (Entered: 08/27/2026)

021

The chilling effect isn't hypothetical. Contractor trade associations and 38 people working in AI told the court the same thing: if a vendor can be blacklisted for what it says, everyone else learns to say less, including about safety risks or concerns about how the government uses its products.

121

The court also rejected the idea that this was merely a dispute over business practices. Citing NRA v. Vullo, it held Anthropic's explanation of its AI redlines, and the practical and moral judgments behind them, was protected speech the government could not punish.

Breadth mattered too. The actions swept far past declining to buy Claude: a government-wide bar, plus a boycott reaching contractors' unrelated business. That's the government acting as a regulator, not as a customer.

131

Evidence of retaliation came largely from the govt itself. Its materials expressly tied its loss of "trust" to the company's public criticism. Hegseth called it a "master class in arrogance," & Trump called Anthropic a “radical left” company. The penalties landed the day after Amodei's statement.

Nothing in this ruling stops DoW from choosing a different model if it wants different capabilities. Agencies have broad discretion over what they buy, and courts rarely second-guess those choices. The problem here was not the choice itself, but the reason for it: Anthropic's public criticism.

122

This is exactly the procurement-leverage problem @cdt.org flagged in its recent report on chatbots and the First Amendment, which noted that Anthropic's allegations, if substantiated, would raise serious First Amendment concerns. cdt.org/insights/who...

Whose Speech Is It Anyway? The Constitutional Contours of Chatbot Regulation

Executive Summary Systems that generate text, images, and audio on demand (or “chatbots”) have become both an important medium of communication and a primary target of regulation. Whether, and on what...

Yesterday, a federal district court ruled that the government violated the Constitution by retaliating against Anthropic — barring it across every federal agency and blocking defense contractors from working with it — because the company publicly criticized the administration on military AI.

David Greene @davidgreene.bsky.social · 26/08/2026

Reposted by Becca Branum

In the Meta settlement, "Age Inappropriate Accounts" includes those that regularly share "Restricted Goods & Services." As we explained in our recent submission to the Meta Oversight Board, that category commonly includes information about abortion medication. www.eff.org/deeplinks/20...

01417

Reposted by Becca Branum

Jaya Saxena @jayasaxena.com · 26/08/2026

The best way to honor the two who have left us recently, imo, is to keep being as gay as possible

184572293

Our argument is simple. The interventions the FTC treats as suspect are what make AI systems accurate, safe, and fair, and they're protected editorial judgment. The FTC can police companies that lie about their products. It can't decide which answers count as neutral.

100

FTC says hallucinations are fine. What draws scrutiny are deliberate choices it dislikes, and it names them: "so-called 'equity,'" correcting historical injustices, following state civil rights law. The trigger isn't falsity. It's what an output says, and whether the FTC agrees.

100

Next up: comments urging the FTC to withdraw its "suppression of accuracy" policy. Calling shaped outputs "deceptive" requires a neutral model to compare against. There isn't one — every model is shaped, at every stage. cdt.org/insights/cdt...

CDT Submits Comments Opposing FTC’s Misguided “Policy Statement Addressing AI Accuracy”

The Center for Democracy & Technology submitted comments urging the Federal Trade Commission (FTC) to withdraw its proposed policy statement regarding the “suppression of accuracy” in artificial intel...

100

Nothing does that reliably. Detectors flag non-native English speakers as bots. Provenance shows how a file was made, not who meant it. So you end up answering the question you can answer — is a real person here? — with ID checks and iris scans. That's a rule about anonymity.

100

If AI Outputs Aren’t Speech, Who Has to Prove They’re Human?

An emerging academic argument suggests that AI outputs aren't "speech," so regulating them raises fewer 1A questions. My worry isn't just the theory, it's the admin. If human writing is protected but machine output isn't, somebody has to tell them apart. www.lawfaremedia.org/article/if-a...

lawfaremedia.org

Denying First Amendment protection to AI outputs sounds like a rule about machines. In practice it could burden human speech online.

100

If AI Outputs Aren’t Speech, Who Has to Prove They’re Human?

Big week at the @cdt.org free expression factory! On the heels of our new report "Whose Speech is it Anyway," two new pieces on how the First Amendment applies to AI outputs & who pays if we get the law wrong. First up: If AI Outputs Aren't Speech, Who Has to Prove They're Human?🧵

lawfaremedia.org

Denying First Amendment protection to AI outputs sounds like a rule about machines. In practice it could burden human speech online.

231

You don't need to love AI to see the danger in letting the government decide what you're allowed to read and which ideas you can share. We can both address real harms from AI and protect users' rights in the process. cdt.org/insights/who...

Whose Speech Is It Anyway? The Constitutional Contours of Chatbot Regulation

052

This isn't to say that AI gets a regulatory pass. Governing what these systems do is urgent and necessary—for our civil rights, our privacy, and more. But when lawmakers target chatbot outputs based on the viewpoints they express, the First Amendment has something to say.

231

Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems

Whether/how the First Amendment protects AI output isn't a theoretical question. FTC just floated treating 'ideological' tuning as consumer fraud. A federal order is pushing AI 'neutrality.' State AGs are probing chatbot 'bias' on political issues. This isn't going away. www.ftc.gov/policy/publi...

ftc.gov

When developers shape chatbot outputs, they're making editorial judgments of the sort courts have consistently protected. But even setting that argument aside, there's an independent basis: your right to share and receive information. Coming or going, the First Amendment protects info via chatbot.

New from @cdt.org: "Whose Speech Is It Anyway? The Constitutional Contours of Chatbot Regulation" on the question suddenly at the center of AI policy: can the gov’t control what a chatbot is allowed to say? Long story short: machines don't have 1A rights, but you and I do.🧵 cdt.org/insights/who...

Whose Speech Is It Anyway? The Constitutional Contours of Chatbot Regulation

Becca Branum @beccabranum.bsky.social · 14/07/2026

This isn’t just a First Amendment problem for researchers. When the government controls which conclusions are acceptable, all of us inherit a weaker body of knowledge. Full comments here cdt.org/wp-content/u...

Becca Branum @beccabranum.bsky.social · 14/07/2026

000

Yesterday @cdt.org filed comments opposing the proposal. Among the red flags: grants terminable at the President's whim, and entire fields at risk of defunding — including research on LGBTQ+ people and anything the gov’t labels “anti-American.”

100

Becca Branum @beccabranum.bsky.social · 14/07/2026

Who decides which research questions are worth being asked? Under OMB’s proposed overhaul of federal grant rules, the answer would be political officials — not scientists, not peer reviewers, and not the merit-based system that has powered American research for decades. 🧵 cdt.org/insights/cdt...

CDT Submits Comments to OMB on Threats to Independent Research

Today, CDT submitted comments opposing the Office of Management and Budget’s proposed overhaul of the rules governing federal financial assistance. The proposal would replace independent scientific ju...

100

Reposted by Becca Branum

Popehat @kenwhite.bsky.social · 21/10/2025

All I am saying is try to live your life so you never have to make those nice young men on Pod Save America look visibly miserable trying to rehabilitate you

402172176

Pitched as protection against deepfakes, the NO FAKES Act would actually create a heckler's veto, and the people best placed to use it would be powerful figures who least like being mocked. Today, @cdt.org led a coalition urging the Senate Judiciary Committee to stop it. cdt.org/insights/cdt...

CDT Leads Coalition Opposing the NO FAKES Act

Today, CDT led a coalition of free expression and digital rights organizations in a letter urging the Senate Judiciary Committee not to advance the NO FAKES Act in its current form. While CDT shares t...

063

And the new likeness right it would create runs up to 70 years after death. Pope Francis passed in 2025, so who could demand this image come down? An Argentine grandnephew? The Vatican? Who knows — but it wouldn't be the man in the coat, who by all accounts may have loved it.

There's a parody exception in the bill, but it's a defense you raise *after* the post is already gone. Platforms facing steep penalties for guessing wrong will just take content down first. The joke disappears while the lawyers sort it out.

122

The NO FAKES Act would create a takedown mechanism for any "highly realistic" AI image of a real person — deceptive or not, harmful or not. Balenciaga Pope is highly realistic. It depicts a real person. That's all the bill requires.

The Pope Francis Puffer Photo Was Real in Our Hearts

Remember Balenciaga Pope? A perfectly serene pontiff, an impossibly puffy coat. A low-stakes, harmless joy of the internet. The NO FAKES Act would put a target on Balenciaga Pope and everything like it. Here's how: 🧵 www.gq.com/story/pope-p...

gq.com

That viral image of the Pope wearing a dripped-out white coat may have been AI-generated, but its swag was the real deal.

132

Reposted by Becca Branum

Center for Democracy & Technology @cdt.org · 08/06/2026

CDT is grateful to have some of our resources on Non-Consensual Intimate Imagery (NCII) reporting processes cited & valued by our partner, Harvard’s Berkman Klein Center, in their latest white paper on survivor-centered reporting systems & policies. More from their resource: asml.cyber.harvard.e...

141

Every filing lands on the same point: the government doesn't get to mark a class of people with a warning label. Pride exists because that kind of erasure has always been worth fighting. LGBTQ+ people belong on screen, and everywhere else.

021

And the breadth of opposition: a 44-org coalition including @hrc.org, @glaad.bsky.social, and others correctly called it a solution in search of a problem. Parents already have the tools. The government has no basis to flag a whole class of people as dangerous. assets.glaad.org/asset/a4b873...

assets.glaad.org

131

This month is Pride Month. At the same time, the FCC is considering whether LGBTQ+ identity in television programming should be treated as a standalone ratings category alongside existing advisories f...

Start with Belle Torek's filing. Her point cuts to the core of the proposal: ratings systems are normative, so making LGBTQ+ identity a warning category signals government disfavor and chills protected speech. www.linkedin.com/posts/bellet...

linkedin.com

ECFS | Belle Torek

122

Advocates push back on proposed FCC warning labels

The FCC is asking whether LGBTQ+ identity should be its own "warning" category in TV ratings, flipping a system built to flag what people do into one that signals disfavor toward who they are. Advocates are rightly pushing back. 🧵 www.washingtonblade.com/2026/06/02/a...

washingtonblade.com

New rating system public notice seeking comments issued on April 22

143

Instead of passing something like the GUARD Act, legislators need to reckon with what users are actually doing when they prompt a chatbot: speaking, listening, creating, learning. The Constitution protects these activities, and any AI regulation needs to take that seriously. cdt.org/insights/thr...

000

The GUARD Act threatens these rights. By targeting tools because they sound "human," demanding ID for routine use, and categorically banning whole modes of conversation for minors, it treats ordinary expression and inquiry as something the government gets to gatekeep.

110

The First Amendment also protects listeners and their right to receive information, explore ideas, encounter perspectives. Using a chatbot to learn or think through a problem are acts of information-seeking that carry constitutional weight.

100

Much like when someone prompts a chatbot, the expressive act can live in the asking: what they choose to inquire about, how they frame it, what they do with the response. Courts have long protected speech that travels through tools and intermediaries. The medium changes; the principle doesn't.

100